Application of Improved Multi-Threshold Birch Clustering in Reservoir Prediction

Porosity is a key indicator for existence of a petroleum reservoir. This paper addresses the porosity estimation problem from seismic impedance volumes and porosity samples located in a small group of exploratory wells. An unsupervised machine learning method was proposed, multi-threshold Birch clustering (M-Birch), showing good performance in clustering of geological facies. Then ridge regression was used to estimate porosity. The results on synthetic data showed that M-Birch method was more accurate than other traditional clustering methods. The results on real data showed the method using M-Birch algorithm combined with ridge regression could obtain reasonable and accurate results of porosity estimation under very sparsely labeled scenario.